\begin{figure}[t!]
	\captionsetup*{justification=centering}
    	\centering 
	
	\caption{Sensitivity Analysis}
	\label{figure-sensitivity-analysis}
   
   	\smallskip 
	
	Panel A: Attended village council meeting last year \\
	
	\smallskip
    	\begin{subfigure}[t]{0.41\textwidth}
      		\includegraphics[width=\textwidth]{./analysis/output/figure-sensitivity-Y-attended-M-yrssch.png}
        		\caption{Years of Schooling}
    	\end{subfigure}
    	~ 
    	\begin{subfigure}[t]{0.41\textwidth}
      		\includegraphics[width=\textwidth]{./analysis/output/figure-sensitivity-Y-attended-M-nchild.png}
        		\caption{No. of Children Living w/ Resp.}
       		 \label{fig:tiger}
   	 \end{subfigure}
    
	Panel B: Discusses politics and community with husband \\
	
	\smallskip
    	\begin{subfigure}[t]{0.41\textwidth}
      		\includegraphics[width=\textwidth]{./analysis/output/figure-sensitivity-Y-discuss-M-yrssch.png}
        		\caption{Years of Schooling}
    	\end{subfigure}
    	~ 
    	\begin{subfigure}[t]{0.41\textwidth}
      		\includegraphics[width=\textwidth]{./analysis/output/figure-sensitivity-Y-discuss-M-nchild.png}
        		\caption{No. of Children Living w/ Resp.}
   	 \end{subfigure}

	Panel C: Participates in a political organization\\
	
	\smallskip
    	\begin{subfigure}[t]{0.41\textwidth}
      		\includegraphics[width=\textwidth]{./analysis/output/figure-sensitivity-Y-member-M-yrssch.png}
        		\caption{Years of Schooling}
    	\end{subfigure}
    	~ 
    	\begin{subfigure}[t]{0.41\textwidth}
      		\includegraphics[width=\textwidth]{./analysis/output/figure-sensitivity-Y-member-M-nchild.png}
        		\caption{No. of Children Living w/ Resp.}
   	 \end{subfigure}
	 	 
        \captionsetup*{justification=justified}

   	\caption*{\footnotesize \textit{Notes:} 
	This figure shows how the estimates of the Average Causal Mediation Effect (ACME) change with different values of $\rho$, defined as the correlation between the error terms in the regression of the mediator on the treatment, and the regression of outcomes on the treatment and the mediator. The dashed line represents the estimated ACME for the given mediator and for different values of $\rho$, while the shaded area represents the 95\% confidence interval. The figures on the left consider years of education as the mediator. Those on the right use the number of children living with the respondent as the mediator. The treatment variable is menarche age. Sequential ignorability implies that $\rho$ is equal to zero, so the ACME for $\rho = 0$  in the above figures corresponds to the ACME estimate in Table \ref{table-mediation-analysis}.}
\end{figure}
